A robust statistical-based estimator for soil moisture retrieval from radar measurements

Abstract

The authors examine the use of a robust statistical inversion approach to the estimation of soil moisture and roughness statistics from backscatter measurements. Two sets of basis functions are examined; the first is a set of basis functions from multinomial combinations of the inputs (termed the MBF) while the second is a set of basis functions generated by a multilayer perceptron referred to as MLPBF. The authors discuss potential sources of training patterns upon which to base these estimators, including empirical forward models and more rigorous theoretical scattering models such as the IEM. These estimators are applied to a set of measured POLARSCAT data from Oh et al. [1992]. Comparisons are made with other inversion methods including neural networks.